Drug–drug–gene interaction risk among opioid users in the U.S. Department of Veterans Affairs : PAIN

Journal Logo

Research Paper

Drug–drug–gene interaction risk among opioid users in the U.S. Department of Veterans Affairs

Chanfreau-Coffinier, Catherinea,*; Tuteja, Sonyb,c; Hull, Leland E.d,e,f; MacDonald, Sallyf,g; Efimova, Olgaa; Bates, Jillh,i; Voora, Deepakh,j; Oslin, David W.b,c; DuVall, Scott L.a,g; Lynch, Julie A.a,f

Author Information
doi: 10.1097/j.pain.0000000000002637

1. Introduction

Opioid overprescribing is associated with misuse, overdoses, fatalities, and overall mortality.3 In 2012, approximately 25% of Veterans seen in outpatient clinics in the Veterans Health Administration (VHA) were receiving an opioid analgesic.14 To address opioid overuse, the VHA launched the Opioid Safety Initiative in 2013. Rates of overall opioid prescribing have since declined; however, the rate of new opioid prescriptions has remained stable over that timeframe.15 Among Veterans, opioid use for chronic pain is accompanied with high rates of psychiatric comorbidities such as substance use disorder and posttraumatic stress disorder.13,14 To alleviate dependence on opioid medications, a multimodal approach to chronic pain treatment is advocated, including the use of adjunctive pain medications such as antidepressants.14,31 Mitigating the risk of adverse events for individuals on opioid therapy is therefore of great importance. One risk reduction strategy would be to understand how factors such as comorbid conditions, drug–drug interactions, and genetic polymorphisms may affect patient response to pharmacotherapy for pain management.

Hepatic cytochrome P450 2D6 (CYP2D6) is a major metabolic enzyme responsible for the elimination of 25% of medications on the market including opioid medications such as codeine, tramadol, and hydrocodone.7,20 It is also the most polymorphic CYP enzyme, resulting in significant interpatient variability in pharmacokinetics, pharmacodynamics, and clinical outcomes for CYP2D6-metabolized drugs.20 Variations in the CYP2D6 genotype can be assessed through pharmacogenetic (PGx) testing to predict how an individual will metabolize drugs processed by CYP2D6 based on classification into 4 phenotypic groups: poor metabolizer (PM), intermediate metabolizer (IM), normal metabolizer (NM), and ultrarapid metabolizer (UM). CYP2D6 phenotype is relevant to the analgesic effectiveness of multiple opioid medications that require bioactivation by CYP2D6 to a more active metabolite that binds the μ-opioid receptor with greater affinity than the parent drug.7 CYP2D6 PMs and, to some extent, CYP2D6 IMs are unable to make this activation and will have decreased therapeutic effect. By contrast, CYP2D6 UMs have increased conversion to the active metabolite resulting in higher maximum blood concentrations and therefore greater incidence of life-threatening side effects such as respiratory depression. In addition to this drug–gene interaction, CYP2D6 is subject to inhibition by commonly prescribed medications resulting in clinically relevant drug–drug–gene interactions and serious adverse reactions; use of a strong inhibitor can even result in a mismatch of genotype and phenotype for a patient, a process called phenoconversion.24 For example, in the presence of paroxetine, a strong CYP2D6 inhibitor, a CYP2D6 NM would have the phenotype of a CYP2D6 PM.

Combining PGx testing of CYP2D6 variants with consideration of CYP2D6 drug–drug–gene interactions may provide an enhanced precision medicine approach to improve the likelihood of successful pain management outcomes. To evaluate the potential impact of PGx testing on pain management among Veterans, we examined the prescribing patterns of CYP2D6-metabolized opioid analgesics, interacting drugs, and adjunctive pain medications in the VHA. We modeled the potential impact of PGx testing on predicting drug response and clinical outcomes and explored how early adopters of PGx testing have used test results for tailoring pain medications in clinical care.

2. Methods

This study was approved by the Salt Lake City VHA and University of Utah Institutional Review Boards and Research and Development Committee and received waivers of informed consent and Health Insurance Portability and Accountability Act authorization.

2.1. Data source

We used electronic health record (EHR) data from U.S. Department of Veterans Affairs (VA) national Corporate Data Warehouse (CDW) transformed into the Observational Medical Outcomes Partnership (OMOP)22 common data model for demographics and outpatient and inpatient medications dispensed by VHA pharmacy; VHA healthcare utilization extracted from CDW Outpatient and Inpatient care data; genetic testing from the VA Genetic Diagnostic test (GDx) data set; and chart content abstracted from the VA electronic medical record called Veterans Health Information Systems and Technology Architecture.

2.2. Study design

We conducted a retrospective analysis for a cohort of VHA patients with at least 1 prescription for an opioid with an established drug–gene interaction with CYP2D6, as determined by the Clinical Pharmacogenetics Implementation Consortium (CPIC, cpicpgx.org) from VA pharmacy in fiscal years 2012 to 2017 (ie, October 1, 2011, to September 30, 2017). These include codeine, hydrocodone, and tramadol based on the current CPIC guidelines.7 Oxycodone was excluded from the analysis, and we limited its inclusion to a sensitivity analysis (Supplemental Material, available at https://links.lww.com/PAIN/B608). Although oxycodone was covered by the CPIC guidelines during the period covered by our study, the level of evidence for an interaction of CYP2D6 with oxycodone was reassessed as lower in 2020 update, and there are currently no recommendations for the dosing of oxycodone based on the CYP2D6 genotype.8

The denominator for the study included all VHA pharmacy users with at least 1 medication record in fiscal years 2012 to 2017 using the OMOP Drug Exposure table, a curated data set for all medications dispensed through VHA pharmacies in outpatient and inpatient settings. The sample includes all VHA pharmacy users with ≥1 opioid prescription in that period (N = 2,436,654; Supplemental Figure 1 for CONSORT chart, available at https://links.lww.com/PAIN/B608). To align with current VA/DoD clinical practice guidelines for chronic pain that are focused on noncancer pain management,33 we excluded patients with an indication of cancer treatment up to a year before the study period based on any occurrence of an ICD-9 or ICD-10 diagnosis code for cancer in the OMOP Condition Occurrence table, any use of an oncology agent using the OMOP Drug Exposure table, or any cancer diagnosis recorded in the VA National Cancer Registry in fiscal years 2011 to 2017. We also excluded patients receiving opioids from cough medications alone.

2.3. Measures

2.3.1. Medication exposure

Opioid prescriptions included all formulations containing codeine, hydrocodone, or tramadol, and an indicator was built to identify cough medications. For each opioid, we built episode of patient exposure using consecutive prescriptions following the guidance of published studies. Two prescriptions were assigned to the same exposure episode if they were separated by less than 180 days and to different episodes if they were separated by 180 days or more.34 Overlapping prescriptions with more than 10 days overlap, or more than 20% overlap, were treated as a single prescription; if the overlap was less than 10 days, or less than 20%, we treated the 2 prescriptions as separate and shifted the end date of exposure by adding the days of overlap. Episode length, total days-supply, and number of prescriptions were calculated for each episode, and exposure episode was classified into short-term use (less than 90 days), long-term episodic use (at least 90 days but less than 10 prescriptions and less than 120 total days-supply), and chronic use (at least 90 days, with at least 10 prescriptions or 120 total days-supply).9,12

2.3.2. Coprescriptions of opioids and other medications

For each patient, we pulled all prescriptions for nonopioid medications of interest overlapping with an episode of opioid exposure (medication list in Supplemental Table 1, available at https://links.lww.com/PAIN/B608). We build a set of indicators for coexposure to a nonopioid medication (yes/no) at the level of the opioid exposure episode. Drugs were classified as strong or moderate inhibitors based on the FDA table of substrates, inhibitors, and inducers.32

2.3.3. Patient characteristics and healthcare utilization

Patient characteristics included sex, race, and ethnicity collected from the OMOP Person table. Race and ethnicity were combined into 4 exclusive categories: Hispanic, non-Hispanic Black, non-Hispanic White, and others. Race and ethnicity are included in the study given differences in variant frequencies observed between groups of different genetic ancestry as well as well-documented differential patterns in pain medication prescribing by race and ethnicity. Age was calculated at the start of the study period (ie, October 1, 2011), and patients were stratified as younger than 65 years vs 65 years and older.

We documented VHA care utilization in fiscal years 2012 to 2017. Primary care (PC) visits, including visits in women's health clinics and geriatric PC, were identified in the CDW Outpatient Visit table using stop codes [301,322,323,350] and counted within each fiscal year; an indicator was built for having 10 PC visits or more during that year. Other patient/year indicators included having at least 1 inpatient stay, at least 1 emergency department (ED) visit, or at least 1 surgery based on the existence of at least 1 record in the OMOP Visit Occurrence table for ED visits and inpatient care, or in the CDW Surgery table. We then cross-referenced the fiscal year of opioid exposure with the year of healthcare utilization to identify patients who had more than 10 PC visits, an inpatient stay, surgery, or at least 1 ED visit, within a year where they were exposed to opioids.

2.4. Chart abstraction

We identified patients in the sample who had a record for CYP2D6 testing in the VA Genetic Diagnostic test database, a curated database of laboratory records for genetic and molecular testing ordered at all VA sites. We looked for any test specific for CYP2D6 genotyping or any pharmacogenomic test panel that included CYP2D6, collected result values available in structured data or in test comment, and curated the result provided as genotype or phenotype into phenotype following the CPIC classification.7,8

Among all patients with documentation of CYP2D6 results during the study period, we abstracted chart information for all patients with abnormal CYP2D6 function that may affect opioid response (n = 45 CYP2D6 poor or ultrarapid metabolizers), as well as a random sample of 45 charts of CYP2D6 NMs or IMs with matching by VA site and year of testing. To facilitate the abstraction process, the abstractors were guided by information on medication exposure and test order available from structured data. Chart abstraction was performed by 2 chart abstractors using the Computerized Patient Record System interface with data entry into the REDCap electronic data capture tools hosted at VA.17,18 Research Electronic Data Capture (REDCap) is a secure, web-based software platform designed to support data capture for research studies.17,18 Agreement between abstractors was evaluated on a set of 10 charts, and a 90% interrater reliability was achieved. Information collected from charts included the existence of pain and mental health diagnoses; documentation of opioid use, antidepressants, and medications that are CYP2D6 inhibitors; and type of CYP2D6 test that was ordered (ie, single-gene test or multigene panel), specialty of the ordering provider, and documentation of the test indication, test results, and medication changes after testing.

2.5. Statistical analysis

Patients were classified as chronic opioid users if they had experienced at least 1 chronic episode use over the study period. Group comparisons were performed using a 2-sided Student t test for continuous variables and Pearson chi-square tests for categorical variables. Odds of opioid chronic use were modeled using a multivariate logistic regression adjusted for patient demographics, healthcare utilization, and year of exposure.

We examined the trend over time for the initiation of treatment with an opioid and concurrent prescription of a nonopioid medication of interest with assigning patients to the fiscal year of the first opioid exposure, so each patient contributes to only 1 fiscal year based on the first opioid prescription. We projected the prevalence of CYP2D6 actionable phenotypes using published frequencies5 adjusted for race/ethnicity distribution in the VA patient population. The predicted frequencies of the phenotypes were applied to the number of patients with prescriptions to predict the impact of drug–gene interaction and drug–drug–gene interaction.

Throughout analyses, estimates were considered significant for P values lower than 0.05. Statistical analyses were performed in STATA 15 (Stata Corp, College Station, TX).

3. Results

There were 2.4 million Veterans with ≥1 opioid prescription from October 1, 2012, to September 30, 2017 (Table 1; Supplemental Figure 1 for the CONSORT chart, available at https://links.lww.com/PAIN/B608). The mean age was 56.2 years (SD 15.8). The cohort included 66% non-Hispanic White Veterans, 20% non-Hispanic Black Veterans, and 7% Hispanic Veterans; 34% of the cohort met the definition of chronic opioid use. Compared with nonchronic users, chronic users were significantly more likely to be older, to be White, and to have higher healthcare utilization during the year of opioid prescription (Supplemental Table 2, https://links.lww.com/PAIN/B608). Chronic users were more frequently prescribed hydrocodone and tramadol than nonchronic users (Supplemental Figure 2, available at https://links.lww.com/PAIN/B608).

Table 1 - Characteristics of Veterans Health Administration pharmacy patients receiving at least 1 opioid prescription in Fiscal Years 2012-2017 by the type of opioid use.
N (%) All (2,436,654) Chronic use (837,660 [34.4]) No chronic use (1,598,994 [65.6])
Male (%) 90.4 92.0 89.5
Age, mean (SD) 56.2 (15.9) 58.6 (14.2) 54.9 (16.6)
 Age < 65 (%) 71.3 69.3 72.4
 Age ≥ 65 (%) 28.7 30.8 27.6
Race/ethnicity (%)
 Non-Hispanic White 66.0 69.6 64.1
 Non-Hispanic Black 19.6 17.5 20.7
 Hispanic 6.8 5.6 7.3
 Others 7.7 7.3 7.9
VHA utilization in the year of opioid exposure (%)
 10 PC visits or more 31.5 43.6 25.1
 Inpatient stay 31.3 37.8 27.9
 Surgery 33.0 34.6 32.1
 At least 1 emergency visit 53.9 52.0 54.8
All differences between groups, P < 0.001; race/ethnicity coded as exclusive categories.
PC, primary care; VHA, Veterans Health Administration.

We examined the potential impact of drug–gene interactions and drug–drug–gene interactions for patients exposed to codeine, hydrocodone, or tramadol. Although the number of new prescriptions for the 3 opioids has declined over the period of observation, the coprescribing of opioid drugs with all antidepressants, tricyclic antidepressants, and neuropathic pain agents (eg, gabapentin and pregabalin) has remained stable over time (Fig. 1 and Supplemental Table 3, available at https://links.lww.com/PAIN/B608).

Figure 1.:
Trends in initiation of CYP2D6-metabolized opioid medications and coprescriptions in fiscal years 2012 to 2017. (A) Number of patients starting an opioid medication in each fiscal year. (B) Percentage of patients receiving an opioid and a concurrent prescription among all patients starting an opioid medication in each fiscal year. Opioids with drug–gene interaction include codeine, hydrocodone, and tramadol. Patients are assigned to the year of first opioid exposure and counted once over the whole observation period. Coprescription includes the following medications by class: antidepressants: bupropion, citalopram, duloxetine, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline, and venlafaxine. NSAID, nonsteroidal anti-inflammatory drugs: celecoxib, flurbiprofen, ibuprofen, meloxicam, and piroxicam; TCA, tricyclic antidepressants: amitriptyline, desipramine, imipramine, and nortriptyline; neuropathic pain agents: gabapentin and pregabalin; CYP2D6 inhibitors: bupropion, duloxetine, fluoxetine, and paroxetine.

Twenty-eight percent of all users had concomitantly received opioids and antidepressant medications affected by the CYP2D6 pathway, with a significantly higher coprescription rate (42%) among chronic users (Supplemental Table 4, available at https://links.lww.com/PAIN/B608). Bupropion, paroxetine, and fluoxetine that are strong CYP2D6 inhibitors and known to phenoconvert the CYP2D6 pathway were coprescribed with opioid medications for 12% of all users, and the rate of coprescriptions was significantly higher among chronic users than among nonchronic users (18.7% vs 8.8%, P < 0.0001). In addition, 3.2% of the patients were concomitantly exposed to a CYP2D6 moderate inhibitor with a higher rate of coexposure among chronic opioid users vs nonchronic users (5.4% vs 2.1%, P < 0.0001; Supplemental Table 4, available at https://links.lww.com/PAIN/B608).

Based on the projected prevalence of CYP2D6 variants in the VHA population, a projected 5.4% (n = 132,615) of the cohort would be classified as CYP2D6 PM and at higher risk for poor pain control, and 3.5% (n = 83,945) would be CYP2D6 UM and at increased risk for opioid-related toxicity (Fig. 2; and Supplemental Table 5, available at https://links.lww.com/PAIN/B608). In this cohort, we observed 10.1% (n = 245,150) of patients taking strong CYP2D6 inhibitors in combination with an opioid medication and 2.7% (n = 65,196) taking moderate CYP2D6 inhibitors (Fig. 2). For patients taking CYP2D6 inhibitors, the analgesic efficacy of the opioid therapy will be based on both the CYP2D6 genotype and the strength of the CYP2D6 inhibitor.2,8 Therefore, an estimated 12.8% (n = 310,346) of the cohort are CYP2D6 NM/IM exposed to a strong or moderate CYP2D6 inhibitor and will be phenoconverted into PM at risk for poor pain control from their opioid analgesics. In total, 22% (n = 526,905) of the patients would be predicted to have a change in response from tramadol, codeine, and hydrocodone because of either a drug–gene interaction or a drug–drug–gene interaction (Fig. 2).

Figure 2.:
Anticipated impact of CYP2D6 drug–gene and drug–drug–gene interaction on medication response among patients prescribed opioids. UM: CYP2D6 ultrarapid metabolizer; PM: CYP2D6 poor metabolizer; IM/NM: CYP2D6 intermediate or normal metabolizer; DGI: drug–gene interaction; DDGI: drug–drug–gene interaction. Based on a cohort of 2,436,654 patients exposed to codeine, hydrocodone, or tramadol, where 297,233 received a CYP2D6 strong inhibitor concurrently with a CYP2D6-metabolized opioid, and 79,047 received a CYP2D6 moderate inhibitor concurrently with a CYP2D6-metabolized opioid; combined prevalence for IM/NM: 82.48%.

As CPIC guidelines included oxycodone as a CYP2D6-interacting drug during the observation period, we performed a sensitivity analysis of the time trends including oxycodone prescriptions and observed similarly a decrease in the overall initiation of treatment but a stable trend in coprescriptions (Supplemental Table 6, available at https://links.lww.com/PAIN/B608).

To assess the availability and use of PGx test results in clinical care during the study period, we performed an exploratory chart analysis to document how early adopters of PGx testing may have used CYP2D6 testing in the care of our patient sample. We identified 521 patients who had been tested for CYP2D6 in the cohort (0.02%). To evaluate prescribing actions that resulted from the availability of CYP2D6 test results, we performed a chart review for 90 patients who had a CYP2D6 test ordered as part of clinical care: all 45 patients with actionable phenotype where a change in opioid dose or drug would be indicated (CYP2D6 UM or PM) and a random sample of 45 controls with phenotype where no change would be indicated (CYP2D6 NM or IM) (Table 2). Most (62%) of the tests were ordered by a psychiatrist. A mental health diagnosis was documented for 98% of the patients, whereas a pain diagnosis was documented for 47%. Although test results were found in the EHR for 98% of the patients, an indication for ordering the CYP2D6 test was documented for only 36% of patients. A change in medications after testing was observed for 38% of the patients. Most medication changes after testing were for psychotropic medications (33%), whereas opioid medications were changed in only 4.5% of patients.

Table 2 - Characteristics and prescribing actions observed after a clinical CYP2D6 test.
Patients with CYP2D6 result (n = 90), N (%)
CYP2D6 phenotype
 Normal metabolizer 38 (42)
 Intermediate metabolizer 7 (8)
 Poor metabolizer 34 (38)
 Ultrarapid metabolizer 11 (12)
Age (y) 51.2 (SD 15.0)
Female sex 16 (18)
Non-White race 13 (14)
Pain diagnoses 42 (47)
Mental health diagnoses 88 (98)
 Depression 61 (68)
 PTSD 53 (59)
 Anxiety 36 (40)
 SMI 14 (16)
 Panic disorder 6 (7)
 ADHD 6 (7)
 Others 20 (22)
More than 1 mental health disorder 72 (80)
Opioid use documented, yes 66 (73)
Chronic opioid use 53 (59)
 Codeine 8 (9)
 Tramadol 23 (26)
 Oxycodone 23 (26)
 Hydrocodone 22 (24)
 >1 opioid documented 26 (29)
Opioid + CYP2D6 inhibitor 13 (14)
Opioid + antidepressant 31 (34)
PGx test type
 Single gene 48 (53)
 Panel 42 (47)
Ordering provider specialty
 Psychiatry 56 (62)
 Primary care 12 (13)
 Internal medicine subspecialty 4 (4)
 Others 8 (9)
 Unknown 10 (11)
Indication for test documented 28 (31)
Test results documented in the medical record, yes 88 (98)
Medication changes documented in medical records 34 (38)
 Psychotropics 30 (33)
 Opioid 3 (3)
 Both 1 (1)
*May be on multiple opioids during the time period.
ADHD, attention-deficit/hyperactivity disorder; PGx, pharmacogenomic; PTSD, posttraumatic stress disorder; SMI, serious mental illness (bipolar disorder or schizophrenia).

4. Discussion

This study examined trends in opioid prescribing across the VHA over a 6-year period and potential for drug–gene and drug–drug–gene interactions. Although opioid use has decreased, the use of coprescriptions of antidepressants and interacting drugs had remained steady among Veterans, and a fair number is exposed to the potential of adverse drug events. Veterans requiring chronic opioid use were more likely to have greater utilization of healthcare resources and receive medications that interact with the CYP2D6 drug-metabolizing pathway, which may affect pain control and occurrence of opioid-related adverse events. An estimated 8.9% of Veterans receiving an opioid (n = 216,560) would have an actionable CYP2D6 phenotype (ie, PM or UM) that may result in nonoptimized pain control.29 Knowledge of this genetic information could help improve treatment because there are alternatives that are less dependent on CYP2D6 that could be used instead; yet, PGx testing was infrequently used in the chronic pain population. The few PGx tests ordered during this timeframe were used to guide prescribing of psychotropic medications, and we found few indications of use to tailor pain medications. An integration of discrete PGx results into the EHR along with clinical decision support systems could help maximize the use of PGx testing for future treatments because results from CYP2D6 tests initially ordered for the management of antidepressants could also be applied to optimize pain management. Our findings can help develop a framework for a comprehensive clinical implementation of PGx testing to optimize pharmacotherapy decisions.

We estimated that 8.9% of Veterans receiving opioid therapy had an actionable CYP2D6 phenotype (ie, PM or UM) that may result in nonoptimized pain.30 In addition, 12.8% were receiving a strong or moderate CYP2D6 inhibitor, which results in phenoconversion to a lower metabolizer phenotype. This means that in approximately 22% of Veterans on opioid therapy (n = 526,905), a thorough assessment of drug–gene or drug–drug–gene interactions may improve pain management. Incorporating this assessment to the list of clinical factors recommended by guidelines to individualize treatment, for example, age, physical diagnoses, and mental health history,33 may help further identify patients at higher risk for toxicity because UMs may experience toxicity after normal doses of CYP2D6-metabolized opioids. It may also prevent ineffective treatment for patients with reduced CYP2D6 activity by informing the selection of alternative medications that are not affected by CYP2D6 activity and overall shorten the time to reach effective pain relief. There may be additional benefits to a genotype-guided approach for pain management. For example, in one prospective, randomized-controlled trial in 282 patients undergoing unilateral joint arthroplasty, the group receiving PGx testing for postoperative pain medication selection had an overall decrease in total opioid consumption as shown by a lower morphine milligram equivalents compared with the usual care arm, without compromising pain control.30

Clinical studies have shown that the CYP2D6 phenotype alone is often not enough to predict the clinical effect. In a prospective crossover study of 17 healthy volunteers, the authors measured the differences in the plasma concentrations of 2 CYP2D6 substrates, tramadol and dextromethorphan, between homozygous (CYP2D6*1/*1) and heterozygous (CYP2D6*1/*17) NM subjects during concomitant treatment with a moderate CYP2D6 inhibitor (duloxetine) or a strong CYP2D6 inhibitor (paroxetine).29 The study found that 71% of the heterozygous CYP2D6 NMs and 25% of the homozygous CYP2D6 NMs were phenoconverted to CYP2D6 IMs by duloxetine. In the case of paroxetine, 94% of the heterozygous CYP2D6 NMs and 56% of the homozygous CYP2D6 NMs were phenoconverted to CYP2D6 PMs. This means that both the strength of the inhibitor (weak/moderate/strong) and the residual CYP enzyme activity as determined by specific genotype are important for predicting the clinical effect. Methods for determining a refined phenotype considering both the CYP2D6 activity score and the strength of inhibition based on FDA classification are available.2,6,8,32

Coprescription of opioids with other classes of medication affected by PGx variants in other drug-metabolizing genes was also observed in our study. Forty percent of the cohort was coprescribed an opioid medication with an antidepressant metabolized by CYP2C19 and/or an NSAID metabolized by CYP2C9, medications that are also influenced by actionable PGx variants. This is likely an underestimate of the number of Veterans using this combination of medications, given over-the-counter availability of several NSAIDs. This study reinforces data from a previous study demonstrating that Veterans are frequently prescribed multiple medications that are subject to drug–gene interactions,5 and therefore, clinical testing for a panel of actionable PGx variants may improve pharmacotherapy outcomes.

Despite mounting evidence to support the clinical utility of CYP2D6-guided opioid therapy,25 several practical implementation challenges threaten the ability to capitalize on these PGx data. One such major hurdle for clinical implementation of a PGx panel is storage of results within the EHR in a structured format that is available to other providers to guide future treatment.19 Indeed, on chart review of Veterans who had CYP2D6 results available, most of them including PGx panels were initially ordered to guide antidepressant therapy and were only available in a PDF format, with little evidence of the results being applied to pain management. Appropriate use of panel testing across indications may benefit patient care for other treatments such as pain management and prevent undertreatment and adverse effects from drug interactions. The VHA is incorporating PGx testing into the EHR under the Pharmacogenomic Testing for Veterans (PHASER) clinical program that will provide PGx panel testing for up to 250,000 Veterans at approximately 50 sites.10 As part of this program, storage of PGx test results as discrete elements along with active or interruptive clinical decision support alerts is being implemented for several actionable drug–gene pairs at the time of medication prescribing. The decision support tools integrating PGx information will complement the current VA Medication Order Check Healthcare Application, an enhanced medication decision support that incorporates different domains from the EHR to alert on drug–drug interactions and duplicate orders and performs a maximal dose check. Currently, Medication Order Check Healthcare Application alerts cover the interaction of the CYP2D6 strong inhibitor bupropion with codeine, tramadol, and hydrocodone. However, alerts are not triggered when fluoxetine, paroxetine, or duloxetine are coprescribed with the CYP2D6-metabolized opioids, highlighting the limitation of these systems and ongoing need to maintain the knowledge base.

Interpretation and application of PGx test results is another hurdle for clinical implementation of PGx testing.27 Several studies have surveyed physicians about PGx testing and concluded that there is a lack of knowledge pertaining to interpretation and application of PGx test results, with less than 15% of physicians reporting feeling well informed about PGx testing and its application.16,21,28 In a more recent survey, 26% of the providers felt confident in their ability to use PGx results in prescribing decisions, although 70% indicated that it would be helpful to consult with a pharmacist when ordering and applying PGx results.26 Successful clinical PGx implementation involves a multidisciplinary approach with inclusion of involvement with a pharmacist specialized in PGx and assessment of drug–drug interactions.1,4,11

The results of this study should be interpreted in the context of its limitations. The use of opioid and nonopioid medications was limited to prescriptions documented in the VA pharmacy records and did not include medications dispensed outside of the VA healthcare system. The prevalence of PGx variants in this cohort was projected using data from the 1000 Genomes Project rather than directly genotyped. We did not systematically capture adverse drug events or therapeutic failures because of drug–gene or drug–drug–gene interactions in this cohort. Drug response phenotypes are challenging to capture using electronic phenotyping methods and often require manual review.23 We did attempt to characterize drug response and medication changes in the 90 participants for whom CYPD6 results were available; however, documentation of adverse events was sparse. We did not include patients with cancer in our analysis to align with guidelines for chronic pain that refer specifically to care outside of cancer treatment, but it is likely that the number of opioid prescriptions and coprescriptions would be higher if we included this population. A similar study would be warranted in the cancer patient population.

In conclusions, our results suggest a potential impact of preemptive PGx testing to inform pain management for Veterans because the coprescription of opioids and interacting drugs is frequent among Veterans being treated for chronic pain. Future research should evaluate the need for this approach in clinical practice and changes in patient outcomes as a result. We also found that in the period studied, PGx laboratory tests were infrequently ordered in this population, and early adoption of PGx testing was performed in the management of mental health conditions rather than opioid management. As the interpretation of PGx test results can be complicated and involve many different genes and classes of medications, further studies are needed to better understand provider needs and the optimal design of decision support tools to integrate drug–gene interactions and drug–drug interactions in the treatment decision.

Conflict of interest statement

D. Voora reports the following supports: Genome Medical (speaker), Sanford Imagenetics (advisory board member), OptumLabs (consultant), and Abbott Diagnostics and AstraZeneca (research grants). S.L. DuVall reports grants from Astellas Pharma, Inc, grants from AstraZeneca Pharmaceuticals LP, grants from Boehringer Ingelheim International GmbH, grants from Celgene Corporation, grants from Eli Lilly and Company, grants from Genentech Inc, grants from Genomic Health, Inc, grants from Gilead Sciences Inc, grants from GlaxoSmithKline PLC, grants from Innocrin Pharmaceuticals Inc, grants from Janssen Pharmaceuticals, Inc, grants from Kantar Health, grants from Myriad Genetic Laboratories, Inc, grants from Novartis International AG, and grants from Parexel International Corporation through the University of Utah or Western Institute for Veterans Research outside the submitted work. The remaining authors have no conflicts of interest to declare.

Appendix A. Supplemental digital content

Supplemental digital content associated with this article can be found online at https://links.lww.com/PAIN/B608.

Supplemental video content

A video abstract associated with this article can be found at https://links.lww.com/PAIN/B609.


This work was supported using resources and facilities at the VA Salt Lake City Health Care System with funding from VA Informatics and Computing Infrastructure (VINCI) [grant no. VA HSR RES 13–457]. This research was supported by a VA Office of Research and Development Merit Review Award awarded to D.W. Oslin (SDR 16-348; ClinicalTrials.gov ID: NCT03170362). S. Tuteja is supported by grants from NHLBI K23HL143161 and the Penn Center for Precision Medicine. The views expressed in this article are those of the author(s) and do not necessarily represent the views of the Department of Veterans Affairs.

Data transparency statement: Deidentified data and program codes used in the analysis will be made available to any researcher for the purpose of reproducing the results on request to the authors and signature of a data transfer agreement.


[1]. Arwood MJ, Dietrich EA, Duong BQ, Smith DM, Cook K, Elchynski A, Rosenberg EI, Huber KN, Nagoshi YL, Wright A, Budd JT, Holland NP, Maska E, Panna D, Elsey AR, Cavallari LH, Wiisanen K, Johnson JA, Gums JG. Design and early implementation successes and challenges of a pharmacogenetics consult clinic. J Clin Med 2020;9:2274.
[2]. Borges S, Desta Z, Jin Y, Faouzi A, Robarge JD, Philips S, Nguyen A, Stearns V, Hayes D, Rae JM, Skaar TC, Flockhart DA, Li L. Composite functional genetic and comedication CYP2D6 activity score in predicting tamoxifen drug exposure among breast cancer patients. J Clin Pharmacol 2010;50:450–8.
[3]. Carey CM, Jena AB, Barnett ML. Patterns of potential opioid misuse and subsequent adverse outcomes in medicare, 2008 to 2012. Ann Intern Med 2018;168:837–45.
[4]. Cavallari LH, Van Driest SL, Prows CA, Bishop JR, Limdi NA, Pratt VM, Ramsey LB, Smith DM, Tuteja S, Duong BQ, Hicks JK, Lee JC, Obeng AO, Beitelshees AL, Bell GC, Blake K, Crona DJ, Dressler L, Gregg RA, Hines LJ, Scott SA, Shelton RC, Weitzel KW, Johnson JA, Peterson JF, Empey PE, Skaar TC. Multi-site investigation of strategies for the clinical implementation of CYP2D6 genotyping to guide drug prescribing. Genet Med 2019;21:2255–63.
[5]. Chanfreau-Coffinier C, Hull LE, Lynch JA, DuVall SL, Damrauer SM, Cunningham FE, Voight BF, Matheny ME, Oslin DW, Icardi MS, Tuteja S. Projected prevalence of actionable pharmacogenetic variants and level A drugs prescribed among US veterans health administration pharmacy users. JAMA Netw Open 2019;2:e195345.
[6]. Cicali EJ, Elchynski AL, Cook KJ, Houder JT, Thomas CD, Smith DM, Elsey A, Johnson JA, Cavallari LH, Wiisanen K. How to integrate CYP2D6 phenoconversion into clinical pharmacogenetics: a tutorial. Clin Pharmacol Ther 2021;110:677–87.
[7]. Crews KR, Gaedigk A, Dunnenberger HM, Leeder JS, Klein TE, Caudle KE, Haidar CE, Shen DD, Callaghan JT, Sadhasivam S, Prows CA, Kharasch ED, Skaar TC. Clinical Pharmacogenetics Implementation Consortium guidelines for cytochrome P450 2D6 genotype and codeine therapy: 2014 update. Clin Pharmacol Ther 2014;95:376–82.
[8]. Crews KR, Monte AA, Huddart R, Caudle KE, Kharasch ED, Gaedigk A, Dunnenberger HM, Leeder JS, Callaghan JT, Samer CF, Klein TE, Haidar CE, Van Driest SL, Ruano G, Sangkuhl K, Cavallari LH, Müller DJ, Prows CA, Nagy M, Somogyi AA, Skaar TC. Clinical pharmacogenetics implementation consortium guideline for CYP2D6, OPRM1, and COMT genotypes and select opioid therapy. Clin Pharmacol Ther 2021;110:888–96.
[9]. Dobscha SK, Morasco BJ, Duckart JP, Macey T, Deyo RA. Correlates of prescription opioid initiation and long-term opioid use in veterans with persistent pain. Clin J Pain 2013;29:102–8.
[10]. Dong OM, Bates J, Chanfreau-Coffinier C, Naglich M, Kelley MJ, Meyer LJ, Icardi M, Vassy JL, Sriram P, Heise CW, Rivas S, Ribeiro M, Jacobitz R, Rozelle S, Chapman JG, Voora D. Veterans Affairs Pharmacogenomic Testing for Veterans (PHASER) clinical program. Pharmacogenomics 2021;22:137–44.
[11]. Dunnenberger HM, Biszewski M, Bell GC, Sereika A, May H, Johnson SG, Hulick PJ, Khandekar J. Implementation of a multidisciplinary pharmacogenomics clinic in a community health system. Am J Health Syst Pharm 2016;73:1956–66.
[12]. Edlund MJ, Austen MA, Sullivan MD, Martin BC, Williams JS, Fortney JC, Hudson TJ. Patterns of opioid use for chronic noncancer pain in the Veterans Health Administration from 2009 to 2011. PAIN 2014;155:2337–43.
[13]. Gallagher RM. Advancing the pain agenda in the veteran population. Anesthesiol Clin 2016;34:357–78.
[14]. Gellad WF, Good CB, Shulkin DJ. Addressing the opioid epidemic in the United States: lessons from the Department of Veterans Affairs. JAMA Intern Med 2017;177:611–12.
[15]. Hadlandsmyth K, Mosher H, Vander Weg MW, Lund BC. Decline in prescription opioids attributable to decreases in long-term use: a retrospective study in the Veterans Health Administration 2010-2016. J Gen Intern Med 2018;33:818–24.
[16]. Haga SB, Burke W, Ginsburg GS, Mills R, Agans R. Primary care physicians' knowledge of and experience with pharmacogenetic testing. Clin Genet 2012;82:388–94.
[17]. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009;42:377–81.
[18]. Harris PA, Taylor R, Minor BL, Elliott V, Fernandez M, O'Neal L, McLeod L, Delacqua G, Delacqua F, Kirby J, Duda SN. The REDCap consortium: building an international community of software platform partners. J Biomed Inform 2019;95:103208.
[19]. Hicks JK, Dunnenberger HM, Gumpper KF, Haidar CE, Hoffman JM. Integrating pharmacogenomics into electronic health records with clinical decision support. Am J Health Syst Pharm 2016;73:1967–76.
[20]. Ingelman-Sundberg M. Genetic polymorphisms of cytochrome P450 2D6 (CYP2D6): clinical consequences, evolutionary aspects and functional diversity. Pharmacogenomics J 2005;5:6–13.
[21]. Johansen Taber KA, Dickinson BD. Pharmacogenomic knowledge gaps and educational resource needs among physicians in selected specialties. Pharmacogenomics Pers Med 2014;7:145–62.
[22]. Lynch KE, Deppen SA, DuVall SL, Viernes B, Cao A, Park D, Hanchrow E, Hewa K, Greaves P, Matheny ME. Incrementally transforming electronic medical records into the observational medical outcomes partnership common data model: a multidimensional quality assurance approach. Appl Clin Inform 2019;10:794–803.
[23]. Robinson JR, Wei WQ, Roden DM, Denny JC. Defining phenotypes from clinical data to drive genomic research. Annu Rev Biomed Data Sci 2018;1:69–92.
[24]. Shah RR, Smith RL. Addressing phenoconversion: the Achilles' heel of personalized medicine. Br J Clin Pharmacol 2015;79:222–40.
[25]. Smith DM, Weitzel KW, Elsey AR, Langaee T, Gong Y, Wake DT, Duong BQ, Hagen M, Harle CA, Mercado E, Nagoshi Y, Newsom K, Wright A, Rosenberg EI, Starostik P, Clare-Salzler MJ, Schmidt SO, Fillingim RB, Johnson JA, Cavallari LH. CYP2D6-guided opioid therapy improves pain control in CYP2D6 intermediate and poor metabolizers: a pragmatic clinical trial. Genet Med 2019;21:1842–50.
[26]. Smith DM, Namvar T, Brown RP, Springfield TB, Peshkin BN, Walsh RJ, Welsh JC, Levin B, Brandt N, Swain SM. Assessment of primary care practitioners' attitudes and interest in pharmacogenomic testing. Pharmacogenomics 2020;21:1085–94.
[27]. Sperber NR, Carpenter JS, Cavallari LH, J Damschroder L, Cooper-DeHoff RM, Denny JC, Ginsburg GS, Guan Y, Horowitz CR, Levy KD, Levy MA, Madden EB, Matheny ME, Pollin TI, Pratt VM, Rosenman M, Voils CI, KWW, Wilke RA, Ryanne Wu R, Orlando LA. Challenges and strategies for implementing genomic services in diverse settings: experiences from the Implementing GeNomics In pracTicE (IGNITE) network. BMC Med Genomics 2017;10:35.
[28]. Stanek EJ, Sanders CL, Taber KA, Khalid M, Patel A, Verbrugge RR, Agatep BC, Aubert RE, Epstein RS, Frueh FW. Adoption of pharmacogenomic testing by US physicians: results of a nationwide survey. Clin Pharmacol Ther 2012;91:450–8.
[29]. Storelli F, Matthey A, Lenglet S, Thomas A, Desmeules J, Daali Y. Impact of CYP2D6 functional allelic variations on phenoconversion and drug-drug interactions. Clin Pharmacol Ther 2018;104:148–57.
[30]. Thomas CD, Parvataneni HK, Gray CF, Deen JT, Prieto HA, Pulido LF, Elsey AR, Elwood EN, Starostik P, Gong Y, Fillingim RB, Johnson JA, Cavallari LH. A hybrid implementation-effectiveness randomized trial of CYP2D6-guided postoperative pain management. Genet Med 2021;23:621–8.
[31]. Turk DC, Wilson HD, Cahana A. Treatment of chronic non-cancer pain. Lancet 2011;377:2226–35.
[32]. US Food and Drug Administration. Drug development and drug interactions: table of substrates, inhibitors and inducers. Available at: https://www.fda.gov/drugs/drug-interactions-labeling/drug-development-and-drug-interactions-table-substrates-inhibitors-and-inducers. Accessed June 23, 2021.
[33]. Veterans Affair/Department of Defense Clinical Practice Guideline for Opioid Therapy for Chronic Pain. Version 3.0 2017. Available at: https://www.healthquality.va.gov/guidelines/Pain/cot/VADoDOTCPG022717.pdf. Accessed Nov 9, 2021.
[34]. Von Korff M, Saunders K, Thomas Ray G, Boudreau D, Campbell C, Merrill J, Sullivan MD, Rutter CM, Silverberg MJ, Banta-Green C, Weisner C. De facto long-term opioid therapy for noncancer pain. Clin J Pain 2008;24:521–7.

Pharmacogenomics; Opioid; Cytochrome P450; CYP2D6; Drug–gene interaction; Drug–drug interaction; Phenoconversion

Supplemental Digital Content

Written work prepared by employees of the Federal Government as part of their official duties is, under the U.S. Copyright Act, a “work of the United States Government” for which copyright protection under Title 17 of the United States Code is not available. As such, copyright does not extend to the contributions of employees of the Federal Government.